The present disclosure generally relates to depth sensing, and specifically relates to a depth camera assembly with a variable frame rate.
In a conventional tracking system, full frames of images acquired by a sensor or a group of sensors (whether a scene or a field-of-view is covered by one or more sensors) are read out from the sensor(s) at some defined rate. The read-out of full frames in the conventional tracking system is irrespective of a distance of a tracking object from the sensor(s) or a relative image size of the tracking object. This full read-out approach is inefficient in terms of frame bandwidth between the sensor(s) and a controller that performs processing of the acquired frames. Also, power dissipation at the sensor(s) and the controller can be prohibitively high due to a large amount of data continuously being acquired by the sensor(s) and processed by the controller. Furthermore, the tracking performance of the full read-out approach can deteriorate as the rate of reading the frames does not change with respect to a distance to an object being tracked—which leads to higher noise and uncertainty in localized coordinates of the object(s) of interest.